Dealing with uncertainty lies at the heart of what it means to be human. Every day we use limited information to make decisions on the basis of predicted future events, which are themselves shrouded in uncertainty. While many of the decisions we make are often banal (for example, determining whether to take an umbrella to work in case it rains), some of the decisions have life-changing and long-lasting impact (for example, determining which course of medical treatment should be taken). Although we make tens if not hundreds of such decisions each day, it is still far from clear how humans determine which path of action should be taken in any given instance.

The computer scientist would be tempted to view human decision-making in much the same way as most machine decision makers are programmed--as probabilistic risk managers, with each possible outcome being conveniently assigned a probability to identify the “most probable” path to take. In many ways, this view underpins much of the recent explosion of interest in the field of machine learning; with sufficient data, a probability distribution can be inferred, which then forms the basis of any future predictions. As such, probability becomes the very currency of human thought.

With this book, Ralph Hertwig and his colleagues at the Center for Adaptive Rationality seek to challenge this worldview. Rather than reducing human decision-making solely to risk management, the authors argue that humans make use of a set of cognitive tools they refer to as the mind’s “adaptive toolbox.” The authors present human prediction, inference, and decision-making in terms of a repertoire of psychologically realistic strategies that fit specific environments. That is, humans make predictions, inferences, and decisions on the basis of simple heuristics, search, and the social environment within which they reside.

The book is loosely structured around these key concepts of the adaptive toolbox--heuristics (Part 2), search (Part 3), and the environment (Part 4). Part 5 examines the possible development and evolution of the mind’s adaptive toolbox, while Part 6 provides a brief historical perspective of uncertainty. Most chapters refer to different studies that the authors have undertaken with regard to specific aspects of the adaptive toolbox.

One of the benefits of the book’s structure is that each chapter stands alone and can be read, and largely understood, independently of the other chapters. However, chapter 1 would appear to be essential reading to ground the remaining chapters within the overall purpose of the book. Indeed, perhaps the best approach for tackling this book would be to read the first chapter, which sets out the research goals, and then the final chapter, which provides a brief history of uncertainty. The remaining chapters can then be read in any order (although I did need to keep a thumb in the excellent glossary at the end of the book, to keep up with the new concepts and terms introduced).

So who should read this book? Beyond those interested in cognitive science and decision-making, this book would be of benefit to those interested in artificial intelligence and machine learning. Gaining a better understanding of how we as humans handle uncertainty will undoubtedly help when seeking to enable machines to handle uncertainty. However, I did find that the initial learning curve required to unpack some of the concepts in this book was relatively steep. Nevertheless, I am glad that I persisted, as the book presents an interesting and thought-provoking view on how we as humans approach the decisions that we make daily.